Guangxuan Song

Guangxuan Song

PhD student, expected graduation in 2026

Thank you for visiting my website! I’m always happy to connect—feel free to reach out.

My research interests lie at the intersection of knowledge graphs and large language models, particularly in their application to scientific data analysis in materials science. I also have a growing interest in exploring how these technologies can support medical research. I am currently pursuing my Ph.D. under the supervision of Prof. Dongmei Fu and the co-supervision of Prof. Dawei Zhang at the University of Science and Technology Beijing (USTB).

A guiding principle I live by is: Do good work, and don’t worry too much about the outcome. (但行好事，莫问前程。) I believe that focusing on meaningful efforts will naturally lead to meaningful results.

📝 Publications

Under Review
sym

Bridging the Semantic-Numerical Gap: A Numerical Reasoning Method of Cross-modal Knowledge Graph for Material Property Prediction

Guangxuan Song, Dongmei Fu, Zhongwei Qiu, Zijiang Yang, Jiaxin Dai, Lingwei Ma, and Dawei Zhang

Code and Dataset

  • We proposed NR-KG, a cross-modal knowledge graph framework that jointly models semantic and numerical information for material property prediction.
  • By addressing the challenges of small-sample scientific data, NR-KG achieves significant improvements over state-of-the-art methods and contributes new benchmark datasets to the field.
Under Review
sym

Taylor-Sensus Network: Embracing Noise to Enlighten Uncertainty for Scientific Data

Guangxuan Song, Dongmei Fu, Zhongwei Qiu, Jintao Meng, and Dawei Zhang

  • We proposed TSNet, a novel Taylor series-based network that explicitly models heteroscedastic noise and estimates both aleatoric and epistemic uncertainty in scientific data.
  • TSNet demonstrates superior robustness and predictive performance across various benchmarks, offering a principled approach to uncertainty modeling in AI for Science.
AECE 2025
sym

A Message Passing Neural Network Framework with Learnable PageRank for Author Impact Assessment

Guangxuan Song, Dongmei Fu, and Xiaomeng Wu

  • We developed NPRNet, a neural-enhanced PageRank framework that integrates graph topology with spatial and attribute information to assess author influence more effectively.
  • Our method improves computational efficiency and better reflects current research dynamics, offering a data-driven tool for scholarly evaluation.

  • Corrosion Resistance Prediction of High-entropy Alloys: Framework and Knowledge Graph-Drive

    Guangxuan Song, Dongmei Fu, Yongjie Lin, Lingwei Ma, and Dawei Zhang

    npj Materials Degradation (Accept)

🎖 Honors and Awards

  • 2020, First Prize, International Contest of innovAtioN (iCAN), Calmspoon Stabilizing Tableware
  • 2020, Gold Award, “Challenge Cup” Capital College Student Entrepreneurship Competition, Calmspoon Stabilizing Tableware (“挑战杯”省赛金奖)
  • 2020, National Second Prize, 4th “Changfeng Cup” National College Student Big Data Competition, Customer profiling for logistics data using knowledge graphs
  • 2021, Second Prize, “Maker Beijing” Regional Innovation and Entrepreneurship Competition – Urban Big Data Track, Calmspoon Stabilizing Tableware
  • 2024, Third Prize, China International College Students Innovation Competition – Beijing Regional, “YiWenZhi” AI-powered medical assistant for clinical documentation
  • 2024, Top 100 Teams, “Jingcai Dachuang” Beijing College Student Innovation and Entrepreneurship Competition – Healthcare Track, “YiWenZhi” Intelligent Medical Assistant
  • 2021, Outstanding Graduate Student Leader, USTB
  • 2022, National Scholarship for Doctoral Students
  • 2022, “Top 100 Youth in the Communist Youth League”, Science Star Award, USTB
  • 2025, May Fourth Youth Medal, USTB (highest honor of the Communist Youth League at USTB)

🧾 Patents

  • 2020, Electronic Anti-Shake Tableware (一种电子防抖餐具), Granted, CN110584428B
  • 2021, Entity Label Clustering Method for Materials Knowledge Graph (一种材料领域知识图谱的实体标签聚类方法及装置), Under Substantive Examination, CN202111258392.8
  • 2022, Construction Method for Metallurgical Materials Knowledge Graph (一种钢铁材料学知识图谱构建方法及系统), Under Substantive Examination, CN202210921904.2
  • 2022, Method and System for Operating a Materials Data Processing Platform (一种材料数据处理平台的搭建、运行方法及系统), Under Substantive Examination, CN202211072133.0
  • 2022, Inference Method for Potential Knowledge in Steel Knowledge Graph (一种基于钢材知识图谱的钢材潜在知识推理方法及系统), Under Substantive Examination, CN202210611454.7

💻 Software Copyrights

  • 2022, Book Knowledge Extraction Annotation Software V1.0 (书籍知识抽取标注软件), 2022SR1352750
  • 2024, Parkinson’s Disease Diagnosis and Medical QA System Based on Knowledge Graph V1.0 (基于知识图谱的帕金森病诊断与医疗问答系统), 2024SR062116
  • 2024, Parkinson’s Disease Medical QA Android App V1.0 (基于知识图谱的帕金森病诊断与医疗问答APP软件), 2024SR0419077
  • 2024, Materials Corrosion Database Query Software V1.0 (材料腐蚀数据库数据检索查询软件), 2024SR0704988
  • 2024, Web-Based Sharing Platform for Corrosion Data Dashboard V1.0 (材料腐蚀数据库和信息资源网络化共享平台数据大屏网站), 2024SR0750810
  • 2024, Client Software for Materials Corrosion Data Platform V1.0 (材料腐蚀数据库和信息资源网络化共享平台网站客户端软件), 2024SR0703838
  • 2024, Integrated Algorithms for Intelligent Corrosion Data Processing V1.0 (材料腐蚀数据智能处理算法集成软件), 2024SR0703369
  • 2024, Medical Risk Assessment System Based on KG and LLMs V1.0 (基于知识图谱和大语言模型的病历风险评估系统), 2024SR0600394
  • 2024, Medical Decision Explanation System Based on KG and LLMs V1.0 (基于知识图谱和大语言模型的医疗决策解释系统), 2024SR0632941

📖 Educations

  • 2020.09 - Now, Ph.D in University of Science and Technology Beijing.
  • 2016.09 - 2020.06, B.S. in University of Science and Technology Beijing.

💬 Invited Talks

  • Oct. 2020, Entrepreneurial Journey: Bringing Warmth to Innovation, iCAN Science Innovation Festival, Qingdao, China | [video]
  • Jul. 2022, From Knowledge Graphs to Industrial Knowledge Automation: A Survey, Chinese Control Conference, Hefei, China (online presentation)
  • Sep. 2024, Knowledge Graphs and Deep Learning in Corrosion Science Data Research, 9th Scientific Data Conference, Chengdu, China
  • Oct. 2024, Knowledge and Data-driven Framework for Multi-principal Element Alloys: Automated Knowledge Acquisition, Representation, and Rediscovery, 22nd International Corrosion Congress, Xi’an, China

🔗 Links

I am fortunate to collaborate closely with the following research partners:

  • Zhongwei Qiu. My senior colleague and research mentor, working on human-centric visual perception, multimodal learning, and generative models with applications in healthcare and science.

  • Meijun Wang. My partner working on autonomous driving and computer vision, with a focus on human pose estimation and pedestrian safety.

  • Xin Qin. My partner working on industrial process fault diagnosis and optimization.